1. Technical Field
The present invention relates to a method for demultiplexing a crossbar non-volatile memory.
The invention particularly, but not exclusively, relates to a method for demultiplexing a semiconductor nanometer-scale (sub-lithographic) crossbar non-volatile memory obtained by using a multispacer structure and to a device for demultiplexing a cross-bar non-volatile memory, and the following description is made with reference to this field of application for convenience of explanation only.
2. Description of the Related Art
As it is well known, in the field of memory devices, the need of realizing circuit configurations of more and more reduced sizes is particularly felt.
In particular, in recent years there has been a growing interest in scaling down the size of Flash technology for non-volatile memories. So, in consideration of the fact that a standard non-volatile memory includes one transistor per cell and in order to increase the memory integration density (number of cells per area), the size of the transistor has been reduced photo-lithographically. In fact, the cell size of a non-volatile memory has been reduced around 10 F2 for NOR cells and around 4.5 F2 for the NAND counterpart, F being the so called technological node.
Even if the International Technology Roadmap of Semiconductors (ITRS) in CMOS technology states that the 32 nm technology node will be introduced in 2013, there are some limiting factors for Flash memory scaling. For example, the scalability of the active dielectric layers of the cell, the so called “tunnel oxide”, is limited by the cell charge retention after many writing cycles of the memory. In particular, the stress induced leakage current could cause statistic failures for oxide thickness less than 9 nm. The reading performance become worse when the oxide thickness is scaled to 6-7 nm, as in ECC (Error Code Correction) memories. More details on Flash scalability limits are described in the publication to R. Bes; E. Camerlenghi; A. Modelli; A. Visconti “Introduction to Flash Memory”, IEEE Vol. 91, 4 pages (489-502) (April 2003).
In order to overcome these limitations on Flash memory scalability, different non-volatile memory layouts have been explored in the last ten years, as described in the following publication: J. B. Brewer, V. V. Zhirnov, and J. A. Hutchby, “Memory Technology for the Post CMOS era”, IEEE Circuits and Device Magazine (March/April 2005).
Recent works have shown that crossbar memories are the most suitable memory layouts for obtaining a tera-scale integration level (>10 Gbit/cm2). A crossbar memory comprises a two-dimensional array of a plurality of nanometer-scale devices, each device comprising a cross-point formed by a pair of crossed wires where one wire crosses another, and at least one switch connecting the pair of crossed wires in the cross-point. In this layout, the memory comprises a matrix whose elements are memory units, each of them implementing a logic function of storing information.
Some examples of crossbar memories are described in the U.S. Pat. No. 6,128,214 and in the WO patent application N.03/094171. In the US patent, a switch described as comprising a bi-stable molecular switch and junctions forming either a resistor or a diode or an asymmetric non-linear resistor. In particular, the state of the junction is capable of being altered by the application of a first voltage and sensed by the application of a second, non-destructive voltage.
The WO patent application discloses a nano-scale flash memory comprising drain and source regions in a plurality of approximately parallel first wires; gate electrodes in a plurality of approximately parallel second wires, the second wires crossing the first wires to form an array of nano-scale transistors. Each transistor can be a configurable transistor or a switch memory bit, set by the application of a voltage. The crossed wire transistors can be formed in a crossbar array.
The existing crossbar memories are usually manufactured by lithography tools, but the most advanced techniques are able to produce arrays having pitch of few tens of nano-meters without using electron-beam lithography, but rather controlling layer thicknesses on the nano-meter length scale (NLS) and transforming vertical into horizontal features. This teaching is described in the following publications: N. A. Melosh, A. Boukai, F. Diana, B. Gerardot, A. Badolato, and J. R. Heath, “Ultra High Density Nanowire Lattices and Circuits”, Science 300, 112 (2003); M. D. Austin, W. Zhang, H. Ge, D. Wasserman, S. A. Lyon and S. Y. Chou, “6 nm Half-Pitch Lines and 0.04 μm2 Static Random Access Memory Patterns by Nanoimprint Lithography”, Nanotechnology 16, 1058-1061 (2005).
It is also known that improvements in terms of resolution and minimum future size of non-volatile memories can be achieved using the Multi-Spacer Pattern Technology (MSPT or SnPT), which is a reiteration of the Space Patterning Technique (SPT) and which is improving with respect to the most advanced lithographic techniques in the realization of repetitive nano-metric architectures. The Multi-Spacer Pattern Technology is a technique through which it is possible to regulate the width of a spacer of a predetermined material (horizontal dimension) by changing the thickness of a thin layer or film, of the same material, deposited on a substrate (vertical dimension).
This technique exploits the possibility to control in an extremely precise way the deposited layer thickness, together with the capacity of many materials to uniformly conform to the topography underlying them.
According to this technique, a seed-block having at least one side wall extending perpendicularly to the substrate is provided. Then, a layer of a suitable material is deposited onto the seed and the substrate and, afterwards, a spacer adjacent to the seed-block side wall is obtained by means of an anisotropic etching of the deposited layer.
The capacity to selectively remove different materials allows further spacers and differently complex structures to be obtained, by subsequent controlled depositions and anisotropic etchings.
In practice, according to this technique, it is possible to realize a structure wherein only one dimension, i.e., the depth, depends on lithography, while the remaining two dimensions, i.e., the height and thickness of the spacer, are obtained by controlling the deposited layer thickness, even within a few nanometers.
An example of a nano-metric architecture obtained by a Multi-Spacer Patterning Technology is disclosed in the publication “A Hybrid Approach to Nanoelectronics”, Nanotech. 16 (2005) 1040-1047, to G. F. Cerofolini et al. In this publication, a multispacer structure 1, shown in FIG. 1, results from three repetitions of the Spacer Pattern Technology (S3PT) and comprises a double layer of conductive spacers 2 (in particular, poly-Si) and of insulating spacers 3 (in particular, SiO2). Some realization steps of the multispacer structure 1 are shown in FIGS. 2A-2D, and namely: a) fabrication with the SPT of a first spacer 2 (for example poly-Si) formed on a silicon substrate 4 covered by a field oxide 5, a sacrificial layer having been provided on the field oxide 5 and etched in the form of a seed 6, the edge portion thereof forming the first spacer 2; b) deposition of a conformal layer 7 (for example, SiO2) covering the seed 6 and the first spacer 2; c) anisotropic etching of said conformal layer 7 and fabrication of a second spacer 3 of the material of said conformal layer (SiO2); d) iteration of the steps b) and c). The multispacer structure 1 resulting from above steps comprises a plurality of spacers 2 and 3 (formed, for example, by a double layer of poly-Si/SiO2), whose height is progressively reduced trough slight changes of the SnPT. The final structure is the covered by a final layer, for instance a TEOS layer.
Moreover, examples of cross-bar architectures obtained trough the Multi-spacer pattern Technology are described in the US patent application N. 2006/051946, in the US patent application N. 2006/051919, in which parallel wires are manufactured exploiting the Multi-Spacer Pattern Technology, and in the publication to G F. Cerofolini, D. Mascolo “Strategies for nanoelectronics” appeared in Micr. Eng. 81 (2005) 405-419 (Aug. 8, 2005).
Using SnPT technique, it is possible to arrange two perpendicular arrays of spacers on different parallel planes, obtaining cross structures with cross-point densities of 1010-1011 cm2. Moreover, if the inter-array distances are controlled and kept on the sub Nanometer Length Scale (NLS), then it is possible to use some known functional molecules as a switch between the two arrays, so as to obtain a memory kernel with density of the order of 0.1 Tbit/cm2. Some examples of this methodology are described in the following publication: G F. Cerofolini, D. Mascolo, “Strategies for nanoelectronics” Micr. Eng. 81 (2005) 405-419 (Aug. 8, 2005); “A Hybrid Approach to Nanoelectronics”, in which the technique is adopted for hybrid CMOS-Molecule crossbar memory.
Nevertheless, the real problem of these devices is to access to the functionalized cross-points and, consequently, to read, confirm or modify the information contained in the memory cells hosted by the cross-points. Therefore the construction of fully nano-scale memory and logic array requires a sort of interface to individually address, probe and change the state of the cross-point devices.
Another aspect to consider is that a functionalized nano-scale crossbar can be used for different applications. For example, for a cross-bar architecture comprising one or few molecules bridging two nanometer-sized electrodes, it should be possible to characterize the electrical behavior knowing only which cell is addressed. So, for this kind of application, used for studying the intrinsic conductance of molecules and the interaction with the contacts, it is not important to know neither the crossbar density nor pitch.
Another possible application relates to the use of a functionalized nano-scale crossbar as a memory, as described in the publication: Y. Luo, C. P. Collier, J. O. Jeppesen, K. A. Nielsen, E. Delonno, G. Ho, J. Perkins, H.-R. Tseng, T. Yamamoto, J. F. Stoddart, and J. R. Heath, Chem. Phys. Chem. 3, 519 (2002). In this case, all memory cells are accessed, namely they are sensed and modified, without necessarily knowing their position in the crossbar arrangement. Instead, the memory cells should not be necessarily addressed, namely it is not necessary to sense and to modify a state of a crosspoint device in any specified position within the crossbar arrangement. For this kind of application, in which the memory cells' local density is small (but not so small as to render impossible probing the state of the memory cell with integrated sense amplifiers), a hardware demultiplexing device is used that has a density which is comparable with the memory one.
It is also possible to use a nano-scale crossbar as a sensor. This application is mainly addressed to biology. For instance, as described in the publication: L. Hood, J. R. Heath, M. E. Phelps, and B. Lin, Science 306, 640 (2004), microfluidic and nanotechnologic platforms are candidates for preventative medicine. Moreover, in the publication to G. F. Cerofolini, G. Ferla, and A. Foglio Para, in Giornale di Fisica 23, 863 (1982), it is shown how a suitably functionalized crossbar could able to probe the membrane of cells with a resolution on the Nanometer Length Scale (NLS). So, for this application, differently from the above described, it should be necessary to know the memory cells' address and their spatial definition, for example in order to map exactly the probed device surface, and it should be necessary to know the memory cell's actual density.
A consequence of the above examples of application of a nano-scale crossbar structure is that the strategy for addressing a crossbar arrangement is application driven. In addition, for addressing a crossbar arrangement it is necessary to link the cross-points of a crossbar memory to some external regions lithographically defined. But this operation involves some important logic, topologic and technological problems, such as an appreciable loss of bit density. The 1:1 connection between sub-lithographic structures and lithographic structures necessarily involves integration losses.
Some methods are disclosed to solve these problems. A first method is disclosed in the U.S. Pat. No. 6,256,767. This document discloses a demultiplexer and a demultiplexing method for a two-dimensional array of a plurality of nanometer-scale switches (molecular wire crossbar network). This document describes a diode- or resistor-based demultiplexer which connects N nano-wires with 5[log2(N)] photolithographic micro-wires. Some randomly distributed metal nano-particles should be deposited over the region of intersection between the control and address wires, and this causes a strict control of the related density. Another drawback of this method is that imprecisely localized nano-particles will lead to intermediate values, altering the individuation of the nano-wires being connected.
Another addressing method is shown in the publication by A. DeHon, P. Lincoln and J. Savage in IEEE Trans. Nanotechnol 3, 165 (2003) and in the WO patent N.2005/029498, which are incorporated by reference herein in their entireties. According to the method disclosed therein, individual nano-wires are addressed trough control micro-wires. The method exploits doped nano-wires acting as field effect transistor, thus allowing the integration of logic gates along a nano-wire itself. The method is based on a “modulation doping” technique, consisting in regulating the doping profile of a nano-scale wire along its axis in order that the threshold voltage of the whole nano-wire can be effectively controlled under the application of a voltage (the same happens in a field effect transistor). So the modulation doping allows an address to be built into a nano-scale wire simply modulating opportunely the doping profile along the wire axis. This can be done during the wire growth phase by controlling the concentration of dopants inside the growth atmosphere at appropriate time. In this way each nano-wire can be associated with a code word. An individual nano-wire which is initially in the non-conducting state, will conduct only if all the control regions are field- or voltage-addressed. So, when a coded nano-wire is aligned across a set of micro-wires, the current flow through the nano-wire can be univocally controlled by the voltage applied at the micro-scale level. The nano-wire behaves as logical equivalent of a multi-input AND gate. However, this technique has the drawback that it is not possible to align the nano-wires, which are randomly assembled over micro-wires.
A third method is described in the publication to R. Beckman, E. Johnstorn-Halperin, Y. Luo, J. E. Green, J. R. Heath “Bridging dimensions: Demultiplexing ultrahigh-density nanowire circuits”, Science vol. 310, page 465, 21 Oct. 2005. This last method is similar to the latter described, but exploiting NOR-logic configuration of nano-wires.
All the known techniques described above make the control of cross-point devices formed by the nano-wires possible through the integration of a hardware demultiplexer in the Nanometer Length Scale (NLS). As a consequence, even if these solutions allows each device to be sensed, they have the drawbacks of increasing the device area and of reducing the memory density, causing a big limitation for the scaling of non-volatile memories realized by cross-bar nanometric architectures.